Election 2017: How the Welsh Political Barometer Did

On respect in which the 2017 general election was certainly not dull was the opinion polling in Wales. Our first two Welsh Political Barometer polls of the election showed clear Conservative leads; there was then a dramatic turnaround, with the final three polls indicating that Labour had a distinct advantage.

 

The opinion polls are often judged around the accuracy of their voting intention figures in relation to the final result. So how did our ‘Final Call’ poll, published on the eve of polling day, compare to the actual election result in Wales?

 

The following table shows the percentage support levels in the final pre-election YouGov poll, and the actual election result, for the last three general elections in Wales:

 

 

Labour Conservative LibDems Plaid UKIP Others*
2010 Final Poll 35% 27% 23% 10% 5%
 

RESULT

 

36.2% 26.1% 20.1% 11.3% 6.2%
2015 Final Poll 38% 26% 7% 12% 13% 4%
 

RESULT

 

36.9% 27.2% 6.5% 12.1% 13.6% 3.7%
2017 Final Poll 46% 34% 5% 9% 5% 1%
RESULT 48.9% 33.6% 4.5% 10.4% 2.0% 0.5%

* Includes UKIP for 2010

 

The standard summary measure for the accuracy of a poll compared to the actual election outcome is the Mean Average Error: how far from the correct outcome the poll was, on average, across all the parties. For the YouGov polls in Wales, the mean average error figures are:

 

2010: 1.5

2015: 0.63

2017: 1.45

 

So how well did we do, overall? Of the three general elections in Wales where YouGov have polled, our final poll this time was the furthest away from the actual outcome with regards to the Labour party. However, we were still within the standard three percentage point ‘margin of error’. Moreover, as we saw with the national Assembly election last year, the previous tendency in many Welsh polls to overstate Labour support appears now to be something that YouGov have overcome. If anything, both last year (very slightly) and this year (to a marginally greater extent), Labour support has been under-stated. But on the big picture – Labour being well ahead of the Conservatives, who were in turn a long way ahead of the other parties – the final Welsh Political Barometer poll was very much correct.

 

None of the three final pre-election polls that YouGov have done before general elections in Wales have ever been far out from the final Conservative vote share. There does appear to be a mild tendency – though no more than that – to under-state support for Plaid Cymru. The only significant error in the final poll this year was with regard to UKIP. But unlike in 2015, UKIP did not stand candidates everywhere in Wales in 2017, and some of those telling our poll that they were intending to vote for UKIP may not, in the end, have had the opportunity to do so.

 

The 2017 election saw major differences between some of the main polling companies’ Britain-wide polls in the final weeks of the campaign. That did make me a little nervous, to be frank, about our Welsh Political Barometer poll. I had no doubts about the professionalism of our colleagues at YouGov; as I have said here previously, I have always found them a joy to work with, and determined to spare no effort in getting things as right as we can make them. But when other much-respected companies, like ICM, were suggesting a significantly larger Conservative lead than were YouGov in the Britain-wide figures, it did suggest the possibility that our Welsh Political Barometer was giving a rather misleading picture. Had we been only one of several polls being run in Wales that would have mattered little, but being the only regular Welsh poll places greater influence upon those of us responsible for the Barometer poll. But our colleagues at YouGov did not let us down – well done to them.

Comments

  • Angharad

    One thing that has occurred to me, and I have heard no-one else mention, is this:
    How many people, when asked how they will vote, give the party they most hope will win, then when the election arrives and they know realistically that their party as no chance, vote for the one most likely to beat their opponents, instead?
    Although polls might try to take this into account with weightings, this election has doubtless been quite exceptional in terms of tactical voting, and could account for some of the discrepancies we see here.

Leave a Reply

Your email address will not be published. Required fields are marked *